Machine learning assessment of pathologic response in lung cancer resections after neoadjuvant therapy - IASLC MPR Project - PubMed
4 hours ago
- #machine learning
- #neoadjuvant therapy
- #pathologic response
- Machine learning algorithms were developed to assess pathologic response (PR) in lung cancer resections after neoadjuvant therapy, aiming to improve efficiency and accuracy.
- Models included a convolutional neural network (digital AI) and a Convex Hull algorithm (CHA), trained on manual pathologist annotations of tumor bed area and residual viable tumor.
- PR was calculated as the percentage of residual viable tumor in the tumor bed area, with comparisons made between pathologist average PR (APR) and digital methods.
- Strong correlations were found between APR vs. digital AI (0.97), APR vs. CHA (0.97), and digital AI vs. CHA (0.99), with 100% agreement for major pathologic response (MPR).
- Concordance for MPR showed a kappa of 0.82 between APR and digital AI/CHA, higher in squamous cell carcinoma (kappa 0.92) than non-squamous carcinoma (kappa 0.77).
- Both APR and digital AI demonstrated similar relapse-free survival (RFS) and overall survival (OS) outcomes, supporting the utility of machine learning in PR evaluation for NSCLC.